Hello,

I recently noticed that the parameter transform_alpha in sklearn.decomposition.DictionaryLearning has a default value equal to 1.0 and not equal to alpha (the main regularization parameter). I think it can be misleading, because we expect the sparse code computed by transform to correspond to the regularization parameter that was used to learn the dictionary, and not some random float.

Is there any reason why it is done that way ? If so, maybe it could be made clearer in the doc.

Thank you in advance for your help !